The Swedish transition to equality: income inequality with new micro data, 1862–1970

The growth of inequality in developed countries since the 1980s has directed a great deal of scholarly attention to the distribution of incomes and wealth, with particular focus how inequality has evolved historically. Recent research explores determinants of inequality such as labour markets (Williamson and Lindert 2016), wars and capital destruction, taxation (Piketty 2014), diseases and disasters (Scheidel 2017), and waves of globalization and politics (Milanovic 2016).

The focus of this project is the long-run evolution of the Swedish income distribution. Sweden plays a special role in research on inequality and welfare states owing to its reputation as uniquely equal (Lindbeck 1997; Andersson 2015). Yet a central question that arises is: how deep are the roots of Swedish equality? Until recently, researchers have focussed on changes since the late 1960s, showing that the expansion of the welfare state narrowed the distribution of income (Korpi and Palme 1998). This short timeframe is a direct consequence of the fact that the first Standard of Livings Survey (LNU) only was made in 1968.

A more history-oriented strand of research has examined the income distribution for shorter time spans before the 1970s with recourse to information provided by Statistics Sweden on taxable income tabulated for broad income groups (Benzel 1952; Spånt 1975; Gustafsson and Johansson 2003). Their results, however, are compromised by two shortcomings with regard to the tax data used: (1) the distribution within income ranges is unknown; and (2) the ranges are not organised systematically over time. In response, researchers have resorted to top incomes, since these have been consistently reported over the years. Roine and Waldenström (2008), for instance, show that the share of the upper ten per cent has declined dramatically in Sweden since 1903. Yet top-income shares reveal nothing about distributional changes that affect the bottom 90 per cent, which may have far-reaching negative implications for our understanding of inequality (Garbinti et al. 2017).

We will provide a substantially improved and detailed view of Swedish inequality by looking directly into the original, individual tax returns sent to Statistics Sweden, presently stored at the National Archive (Riksarkivet). We will estimate incomes for the full Swedish population using the social tables method, employed with great success in recent estimates of income inequality in the US, the UK, and Germany as far back as the eighteenth century (Lindert and Williamson 2016, Gomez-Leon and de Jong 2017). The method combines (1) the distribution of the population in sectors, occupations, and social groups from population censuses, and (2) income estimates for each group. It will allow us to provide new and unique decadal estimates of Gini-coefficients, top incomes, bottom incomes, regional differences, gender differences, and urban-rural differences for the entire Swedish population from 1862 to 1970. The estimates for 1970 will be linked up with existing data to produce one of the longest and most qualitative series of income inequality for any country. By providing uniquely rich data on the Swedish income distribution, this project will make a major contribution to the international debate on inequality and its determinants.

Sverige utmärkte sig under efterkrigstiden som ett av världens mest jämlika länder. Hur det blev så vet vi förvånansvärt lite om: tillgängliga data före 1968 berör endast toppinkomsttagare eller enskilda platser. Baserat på ett unikt och tidigare outnyttjat inkomsttaxeringsmaterial presenterar det här projektet den första beräkningen av hela den svenska befolkningens inkomster mellan 1860 och 1970. Statistiken gör det möjligt att utvärdera olika faktorers betydelse för ojämlikhetens utveckling över det långa loppet. Projektet bidrar således med nya insikter om den svenska inkomstfördelningens utveckling och med en förståelse för dess drivkrafter som är unik i ett internationellt perspektiv.